Albert F. Lee, Ph.D.

A man who learns SAS in Statistics, falls in love with R and is crazy about Python

Data Science Applications in Healthcare

Types of questions Data Science can help answer:

1. Classification: likelihood of hospitalization; likelihood of Emergence Department visits; likelihood of medication adherence; likelihood of self-management (motivation index – formerly MEDai of LexisNexis); CMS-HCC; HHS-HCC model in risk adjustment area (suspect logic to maximize revenue)

2. Poisson regression: forecast of inpatient length of stay (LOS); forecast of ED visits in the next 12 months

3. Regression: forecast of total medical cost in the next 12 months; forecast of total inpatient cost in the next 12 months; outpatient cost in the next 12 months, pharmacy cost in the next 12 months;

4. Anomaly detection: Special Investigation unit (SIU) – fraud detection (Fraud, Waste and Abuse); 7 level of fraud, license to Steal in healthcare (Malcolm Sparrow)

5. Quality analytics – Health Effectiveness Data Information Set (HEDIS): Quality incentive and quality withhold, NCQA accreditation, auto-assign measures: Controlling High Blood Pressure (CBP); Cervical cancer screening (CCS); comprehensive diabetes care (CDC); childhood immunization status (CIS); prenatal care (timeliness of prenatal care) PPC; Well-child visit in third, fourth, fifth and sixth years of life (W34) – 37 points out of 50 points – Time series modeling on prospective and retrospective rates, NCQA benchmarks 40-50 hybrid measures that require medical record review chase – within 5 month-time frame (take a random sample, chase medical records in physician’s office for positive hits)

6. CAHPS- consumer assessment of healthplan providers and systems (13 points)

7. Natural Language Processing – NLP: related to medical record chases; Clinical Decision support

I’ll begin the series of blog to share what I have learned in my career

https://leealbert52.github.io